Results and discussion Directory UMM :Data Elmu:jurnal:A:Agriculture, Ecosystems and Environment:Vol81.Issue2.Oct2000:

M.J. Kirkby et al. Agriculture, Ecosystems and Environment 81 2000 125–135 131 Integration of the storm runoff and sediment trans- port equations can be carried out analytically over this distribution to give the form of the erosion risk ex- pression. The overall effectiveness of the erosion risk estimates may be compared with observed empirical relationships between sediment loss and climate, with a minimum in temperate zones. A model compari- son Kirkby, 1995 with a transect across the southern United States shows fair qualitative agreement with empirical summaries Langbein and Schumm, 1958.

3. Results and discussion

3.1. Applications to erosion risk in France A preliminary attempt has been made to imple- ment these methods for France, based on data held by INRA, Orléans, and which has already been used to prepare a preliminary qualitative assessment Montier et al., 1998. Parameters have been derived from the CORINE land cover survey, a 5 km gridded database for interpolated monthly mean rainfall totals, a 250 m resolution DEM and information from the draft Eu- ropean Geographic Soils Database. These have been used to create qualitative pedo-transfer functions, based primarily on INRA experience, to estimate monthly vegetation cover, runoff thresholds, crusting class and soil erodibility. A preliminary map, show- ing the feasibility of the approach, is shown in Fig. 6 Yassoglou and Jones, 1998; Kirkby et al., 1998, but the results are, to date, not fully validated. This exercise demonstrates that these methods are able to lead to reasonable estimates of monthly average erosion rates, but that there is considerable scope for improvement in parameterisation from existing data sources, the creation of additional special purpose data layers within European Databases, and for a substan- tial exercise in validation at a range of appropriate scales. This work has been completed, using the meth- ods set out above, generating monthly and annual estimates of erosion risk. These are presented in cat- egories, which are provisionally classified as mean erosion loss rates. At present pedo-transfer rules have been based on the experience of Le Bissonais and other colleagues at INRA, Orléans, but the final product is un-validated at this stage. Although the resolution shown is 250 m, it should also be recog- nised that less reliance should be made on individual pixel values than on the clear regional patterns. Fore- casts are available for each month, and could also be expressed in terms of the probability of exceedance of a given event size. 3.2. Data sources 3.2.1. Soils The European Geographic Soils Database for France is clearly the best and main source for erodi- bility and some of the soil storage terms. Following the results of Montier et al. 1998 and Le Bissonais et al. 1996, the values were used for the trial run based on data in the European Soils Database Tables 1–3. 3.2.2. Land cover The CORINE land cover map has been used. Stan- dard functional conversions were used to convert cover types to vegetation cover Table 4. Land cover classes were taken from the CORINE database. Additional information for types of arable farming were taken from the table of Petits Régions Agricoles PRA crop returns for the same year, linked Table 1 Erodibility classes from categories in European soils database Relative erodibility class a Value Level of confidence RMS error Weak 1 Weak ± 80 Moderate 3 Moderate ± 50 Strong 10 Strong ± 20 a The relative erodibility class is taken directly from the pedo- transfer functions in Montier et al., 1998. Table 2 Crusting classes from categories in European soils database Relative crusting class a Crust storage b , h C mm Level of confidence RMS error None na Weak 12 Weak ± 80 Moderate 6 Moderate ± 50 Strong 2 Strong ± 20 a The relative crusting class was taken directly from the classes ‘Battance’ in Montier et al., 1998. b The crust storage is greater than the total soil storage, then it is replaced by the soil storage, h s + h R . 132 M.J. Kirkby et al. Agriculture, Ecosystems and Environment 81 2000 125–135 Fig. 6. Preliminary erosion risk map of France, at 250 m resolution. Table 3 Water retention estimates from categories in European soils database Soil texture class a Water retention mm at saturation Water retention mm at field capacity 50 cm Soil storage, h S mm Coarse 403 294 109 Medium 439 379 60 Medium fine 430 406 24 Fine 520 472 48 Very fine 614 567 47 Organic soils 766 708 58 a Soil texture class was taken from the soil database. The values have been provided Christine Le Bas, personal communication from an experimental pedo-transfer function. M.J. Kirkby et al. Agriculture, Ecosystems and Environment 81 2000 125–135 133 Table 4 Land cover types from CORINE classification in ESDB Land cover type a Initial surface storage, h R , mm Reduction per month Arable 10 50 Other 5 a Land cover type was also taken from the CORINE database. Arable storage was set to its initial value at times of cultivation taken as seed-time and harvest. Table 5 Interception as percentage of storm rainfall estimated from ESDB Land cover type Interception storage, h I , as of storm rainfall Arable 5 Pasture, vineyards and tree crops 10 Forest 20 Heterogeneous 10 Natural degraded land 5 Urban, rock, wetlands na to the database layer which identifies the PRA of each cell. This gave the proportions of winter and spring sown arable, together with a total arable. If there were areas with both categories of arable, subtraction gave an estimate of the area with both types in the same year. This provided estimates of initial surface storage Table 4, interception Table 5 and vegetation cover for each month of the year Table 6. 3.2.3. Topography Relief was estimated as the standard deviation of re- lief at each point in a DEM, based on all points within Table 6 Land cover type from CORINE database within ESDB Land cover type January February March April May June July August September October November December Arable Winter sown 10 20 40 60 80 100 100 50 10 Spring sown 10 10 10 20 50 80 100 100 50 10 10 Both in 1 year 10 10 10 20 50 80 100 100 50 10 Pasture 100 100 100 100 100 100 100 100 100 100 100 100 Permanent vineyards, tree crops, etc. 30 30 30 40 50 60 60 60 60 40 30 30 Forest 100 100 100 100 100 100 100 100 100 100 100 100 Heterogeneous 50 50 50 60 70 80 90 90 60 50 45 45 Natural degraded 20 20 20 20 20 20 20 20 20 20 20 20 Rock, urban, wetlands, etc. na na na na na na na na na na na na a given radius. Test show that this is not sensitive to the DEM resolution, although the radius used is limited to requiring at least five points in the sample i.e mini- mum radius≥DEM cell size. A radius of 1 km worked well, and was satisfactory using the 250 m DEM for France available, with permission, through INRA. 3.2.4. Climate The 5 km interpolated rainfall map for France gave an excellent quality for monthly mean rainfalls. In addition the table for ≈90 stations in France gave the frequency of 24 h totals which were used to obtain the frequency distribution of daily rainfalls. Values of the mean rain per rain day and its standard deviation were then assigned back to the 5 km grid to fit the distributions of monthly rainfall intensities. 3.2.5. List of data required All of these are held by INRA for France, and are subject to permission for use in the proposed context Table 7. They comprise a series of data layers from the soils database at 250 m resolution, monthly pre- cipitation at 5 km resolution, additional tables of Pe- tits regions Agricoles PRA and station climatic data. These data are then processed using a series of Arc Macro Language AML algorithms. 3.3. Comparison with the INRA erosion risk map of France Given the time available, there has been no statis- tical analysis of differences from the INRA study, but only a visual comparison. The pattern of seasonal differences look realistic, and some areas show good 134 M.J. Kirkby et al. Agriculture, Ecosystems and Environment 81 2000 125–135 Table 7 Data required for LQI erosion model Layers in soil database at 250 m resolution 1. Erodibility classes and reliability 2 layers 2. Crusting classes and reliability 2 layers 3. Soil texture classes 4. CORINE land use classes 5. PRA membership 6. Elevation DEMMNT 7. Slope classes for comparison only Data layers at 5 km resolution Monthly mean precipitation 12 layers Tables 1. PRA: Partition between crop classes esp. winter and spring sown arable 2. Climate data for 90 stations. Distribution of intensities for 24 h rainfall by month if available AlgorithmProgram Interpolation routine used to apply metro station data to 5 km grid. The complete set of rules and algorithms was applied through macros AML in Arc-Info to obtain the final maps for average erosion risk in each month, as an estimated mean erosional loss. The annual map is the sum of the 12 months local and regional convergence with the INRA maps, but there are also divergences. For example, the Laura- gais SE of Toulouse is well known for erosion prob- lems, which are better identified in the RDI than in the INRA map. However, in other areas the RDI model seems to overestimate erosion, such as in central Bre- tagne, between Rennes and Nantes, in Basse Nor- mandie between Caen and Granville, and in the Jura area along the Swiss border. Many of these areas are almost completely covered by vegetation, grassland and forest, respectively. In the initial runs, the whole of south and east France generally also showed too high an erosion rate, and this contrast has been reduced by modifying the relief factor in the RDI estimator. It is planned to improve and validate the model properly within a Framework V research grant PESERA.

4. Conclusions